For hosts
Browse by Person — the guest list, assembled from the photos
A People tab beside the gallery grid, built from the faces already found during indexing. No names, no tagging, no address book.
A gallery with a few thousand frames has a few dozen recurring people in it, and no practical way to answer "which ones is Meera in?". Tagging solves it in theory and never in practice, because tagging is work someone has to do after an event everybody has already moved on from.
Browse by Person does it from the faces themselves. Every face the indexer found is grouped by similarity into a person, each person gets a cropped circle, and opening one lists that person's photos newest-first. It is deliberately a different, stricter grouping than face search: search is answering "is this the same person as this selfie", clustering is answering "who keeps appearing", and the second needs a tighter threshold to avoid merging two guests into one tile.
Someone who appears in only one photo is left out on purpose — a single face is a coincidence, not a person worth a tile.
How it works
- 1
Faces are found during indexing
The same pass that makes an event searchable also records every face it detects.
- 2
Build the people list
Faces are grouped by similarity into people. It runs on request rather than automatically, so a gallery still receiving uploads isn't clustered halfway.
- 3
Open a person
Their photos load newest-first, sorted by capture time. Back returns you to all people.
- 4
Rebuild after more uploads
New photos mean new faces — rebuilding folds them into the existing people.
At a glance
- Where
- People tab, inside the event gallery
- Who can open it
- Event owner, editors and admins
- Minimum to form a person
- 2 separate photos
- Photos per person
- 60 at a time, up to 200
- Face crop
- 256 px square, from their sharpest frame
Availability
Rolling out behind a feature flag, so it appears on an event only once it is switched on. Owner, editor and admin only.
Questions
- Do guests see the People tab?
- No. It is a management surface — the event's owner, its editors and admins. Guests find their own photos with face search instead, which only ever returns photos of the person who submitted the selfie.
- Why is someone missing from the list?
- Three reasons account for almost all of it: they appear in only one photo (a person needs two separate frames), their face was too small or too turned away to score confidently, or they have used the face opt-out for this event.
- Is this the same thing as face search?
- No, and the difference is deliberate. Face search matches one selfie against the event. Clustering groups the event against itself, using a stricter similarity threshold — a loose one would quietly merge two guests who look alike into a single tile.
- Someone was grouped wrongly. What now?
- Rebuild the people list. Clustering is derived data, so rebuilding recomputes it from scratch against the current set of photos — it never edits or deletes a photo.
For hosts — more
- Photos & VideoStills and clips in one gallery, in capture order — and face search reads both.
- Timeline & MapThe event replayed in capture order and pinned to where it happened — read from each photo's own EXIF.
- Album Photo SelectionPick the album shots on a screen built for picking, and send the set straight to the photographer.
- CollectionsCut one gallery into named sections — Mehendi, Ceremony, Reception — without duplicating a single file.
- Download everythingZip the whole gallery into one file, built on our side so your device isn't asked to hold the event.